Professional Documents
Culture Documents
“Normal” air temperature can have differing meanings. The World Meteorological
Organization attempted to formalize normal temperatures in order to have a standardized
international period. This helps in determining climatological trends. At the International
Meteorological Conference in Warsaw, Poland in 1935, the years 1901-1930 were selected as the
international standard period for normals. In the U.S., NOAA’s (National Oceanic and
Atmospheric Administration) National Weather Service (previously the Weather Bureau)
adopted this 30-year standard and now, recalculates “normals” at the end of each decade, using
the preceding 30 years’ data as input. This practice accounts for slow changes in climate and
adds data from more recently established stations to the network. When computing “normals” for
a 30-year period, as with an annual temperature trace, trends from warmer to cooler, and warmer
again are seen. However, this line is not a smooth line, as shown in the following figures (Figs. 1
and 2). A statistical technique is used to smooth out the line to eliminate daily “jumping” of
values. The information presented here does not employ the smoothing technique. Additionally,
this paper argues that in extreme climates, such as Montana, the term “average temperature”
should be used, rather than “normal temperature.”
115
95
Temperature (F)
75
55
35
15
-5
-25 J1 10 20 F1 10 20 M1 10 20 A 1 10 20 M 1 10 20 J 1 10 20 J 1 10 20 A 1 10 20 S 1 10 20 O 1 10 20 N 1 10 20 D 1 10 20 30
Date
The largest variation occurs during the winter, with a standard deviation of 22 on January 10, 11
and December 29. The lowest standard deviation occurs in the summer, with the lowest of 6°F
on July 28.
Figure 2. Minimum Temperature at Great Falls (1971-2000):
Period Extreme Minimum, Mean temperature, Lowest Minimum and one Standard Deviation.
75
55
Temperature (F)
35
15
-5
-25
-45 J1 10 20 F1 10 20 M1 10 20 A 1 10 20 M 1 10 20 J 1 10 20 J 1 10 20 A 1 10 20 S 1 10 20 O 1 10 20 N 1 10 20 D 1 10 20 30
Date
The largest variation again occurs during the winter, with a standard deviation of 22 on January
11. The lowest standard deviation occurs in the summer, with the lowest of 4°F on July 28.
__________________________________________
In computing a “normal” or average temperature, daily values for the preceding 30 years
are used. Because of daily temperature fluctuations, this “normal” value likely includes several
values not even close to the “normal.” To express the spread, or variance, in the computations,
we can employ other statistical functions by looking at a normal distribution of the data and
finding values such as a standard deviation. In a normal distribution, 68% of the 30 values (20 of
the 30) will fall within one standard deviation of the computed average temperature. Simply put,
the standard deviation is an expression of the scattering, of the observations. If one standard
deviation is small (4°F), the “normal” or average temperature is a good representation of the
input values. So, at Great Falls in July, an average minimum temperature of 56°F would be a
good representation of all 30 values. If the standard deviation is large (15-25°F), the “normal”
value has much more scatter associated with it, so more likely implies that it is just an average of
the values, rather than a “normal” value. As shown in Figures 1 and 2, the largest variation in
maximum temperature is in January. Looking at January 11, the range in maximum values is
59°F to -17°F! The range in minimum values is 44°F to -29°F. These ranges underscore the wide
fluctuation in values used in computing the “normal.” In July, on the 28th, the lowest standard
deviation occurred. On this date, the range in high temperatures was 100°F to 72°F. The range in
minimum values was 64° to 45°F, or much smaller than the range in temperatures in January.
This variation becomes important in Montana when “normal” temperatures are discussed
in the cool season. In the summer months, the standard deviation of the temperature (both
maximum and minimum) is generally less than 10°F. In the cool months, the standard deviation
reaches more than 20°F. While NOAA’s National Weather Service releases climatological
information relating to the computed “normal,” in reality, if the standard deviation is 20°F with a
“normal” of 30°F, there is as much of a chance that the temperature could be 15°F, 30°F, or even
45°F! A daily probability chart could show this. On a daily basis, this may be noteworthy, but it
is only over a longer period of time, over a week or more, where the significance of above (or
below) average values bears out.
Figure 3. Cut Bank October 2002 Minimum Temperatures Departures from Average.
10
Departure from Average
0
10
13
16
19
22
25
28
31
1
4
7
Departure from
-10
Average
-40
Date
There are other ways to look at the importance of an average, rather than “normal.” In
another example, at Great Falls in January 1996 (Fig. 4), the month’s average temperature was
11°F below “normal.” The low temperatures averaged 13°F below “normal.” On a daily basis,
the variation was quite dramatic as shown in the following figure. There were only five days in
which the minimum temperature was within five degrees of the daily “normal” temperature, and
only three days within three degrees of the daily “normal.” With standard deviations between 15-
22°F during the month, fully one-half of the month’s temperatures fell within one standard
deviation of the computed “normal.” A frequency distribution of ranges of average temperature
at Great Falls (Fig. 5) further emphasizes these deviations. This shows, that even though the
“normal” temperature is around 22°F, this only occurred about 10 percent of the time in the 30
year period.
Figure 4. Great Falls January 1996 Minimum Temperature Departures from Average
GTF Jan 1996 Minimum Temperature
Departure from Average
30
20
Departure (F)
10
0
-10
-20
-30
-40
-50
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31
Day
The greatest occurrence of values was actually between 30-32°F. These values occurred 20
percent of the time. This further emphasizes that the computed normals are not really normals,
but averages, and should be so stated.
15
Frequency (%)
10
0
3 6 9 12 15 18 21 24 27 30 33 36 39 42
Temperature (F)
Another example uses mean January temperatures for Great Falls from 1971 through
2000. Figure 6 shows that even the monthly averages show wide variations. The standard
deviation for this period is 8°F, meaning that the monthly mean temperature, or “normal” may be
anywhere from 14.8°F to 30.8°F! Indeed, during this 30-year period, only seven or 23% of the
values were within three degrees of the “normal value. Thirteen other years’ values fell within
the one standard deviation interval of 8°F.
20
Departure from
10
Average (F)
0
1970 1975 1980 1985 1990 1995 2000
-10
-20
Year (Mean 22.8F)
While NOAA’s National Weather Service will likely continue to publish “normal”
temperature values, one needs to keep in mind that the “normal” is simply an average of all the
values. It is possible that not even one of the input values is equal to the computed “normal,” and
the most commonly occurring values may be significantly separated from the “normal”. This
further underscores that, although it is interesting to know the “normal” or average temperature;
it is probably more important to understand the distribution of the values used in computing the
“normal,” which the ranges, standard deviation, variance and frequency distributions can relate.
References
Cooper, D. & Schindler, P. 2000. Business Research Methods, 7th Ed. McGraw Hill Irwin.
Boston.
Crutchfield, H. 1983. General Climatology, 4th Ed. Prentice Hall, Englewood Cliffs, NJ.
Levin, R. 1981. Statistics for Management, 2nd Ed. Prentice Hall, Englewood Cliffs, NJ.